I have been looking for the same thing, either from Meta's SAM 3[1] model, either from things like the OP. There has been some research specifically in this ar

I have been looking for the same thing, either from Meta's SAM 3[1] model, either from things like the OP. There has been some research specifically in this area with what appears to be classic ML models [2], but it's unclear to me if it can generalize to dances it has not been trained on. [1] https://ai.meta.com/blog/segment-anything-model-3/ https://ai.meta.com/blog/segment-anything-model-3/ [2] https://arxiv.org/html/2405.19727v1 https://arxiv.org/html/2405.19727v1